Image edits

POSThttps://api.inferencebureau.com/v1/images/edits

Change an existing image from a text prompt, with an optional mask for inpainting, in the OpenAI images shape. Admitted, held, settled and billed like image generation, on the models certified for edits.

Body

application/json, or multipart/form-data as OpenAI's SDKs send it; up to 4 MiB

modelstringrequired
An image model certified for edits. GET /v1/models marks them with edits: true; others are refused with 400 invalid_request.
promptstringrequired
What to change, up to 32,000 characters.
imagesarrayrequired
One to four input images, each { "image_url": … }: an https URL the provider fetches, or a base64 data URL of a PNG, JPEG or WebP image. In a multipart form, send them as image or image[] files. Large images are better sent as URLs, to stay under the request limit.
maskobjectoptional
Where to edit, as { "image_url": … } or a mask file: transparent pixels mark the area to change. Inpainting models such as bfl/flux-pro-1.0-fill need one.
n, size, response_format, providerOptions, user, iboptional
As for image generation. In a multipart form, send providerOptions as a JSON string.

Notes

Costoptional
Held and settled as for generation. On token-priced models the input images are billed too: each is held at the most input tokens one took when we measured it at the highest input fidelity, priced at the output rate, which is above what providers charge for image input.
Logsoptional
When your team keeps request content, the stored request shows each input image's size, not the image.